Anthropic Challenges Pentagon’s Claims of National Security Risks
Late Friday, Anthropic submitted two sworn declarations to a federal court in California, countering the Pentagon’s assertion that the AI company poses an “unacceptable risk to national security.” The declarations argue that the government’s case is built on technical misunderstandings and unfounded claims that were not raised during the preliminary negotiations.
Legal Proceedings and Impending Hearing
The filings were made alongside Anthropic’s reply brief in its ongoing lawsuit against the Department of Defense. This legal skirmish is nearing a critical moment, with a hearing scheduled for Tuesday, March 24, before Judge Rita Lin in San Francisco.
Background of the Dispute
This dispute traces back to late February, when then-President Trump and Defense Secretary Pete Hegseth publicly announced the decision to sever ties with Anthropic. The company refused to grant unrestricted military access to its AI technology, a move that ignited tensions.
Expertise of Declarants
The declarations come from Sarah Heck, Anthropic’s Head of Policy, and Thiyagu Ramasamy, the company’s Head of Public Sector. Heck, a former official of the National Security Council who has held positions at the White House during the Obama administration and at Stripe, manages Anthropic’s government relations and policy initiatives. She was present at a key meeting on February 24, where CEO Dario Amodei discussed issues with Defense Secretary Hegseth and Under Secretary Emil Michael.
Falsehoods in Government Filings
In her declaration, Heck challenges what she characterizes as a principal falsehood in the government’s documentation: the claim that Anthropic sought a formal approval role regarding military operations. She firmly states that no Anthropic employee articulated such a desire during negotiations. Furthermore, she points out that the Pentagon’s fears regarding Anthropic potentially disabling or modifying its technology mid-operation were never addressed in discussions, only surfacing later in court filings.
Contradictory Communications
One notable detail from Heck’s declaration involves an email sent by Under Secretary Michael to Amodei on March 4, shortly after the Pentagon formally categorized Anthropic’s technology as a supply-chain risk. In this correspondence, Michael indicated that the two parties were “very close” to aligning on controversial issues related to autonomous weapons and the surveillance of American citizens. Heck implies a contradiction: if these issues indeed make Anthropic a national security threat, why was there a perception of alignment shortly after the risk designation?
Technical Merits of Anthropic’s Technology
Ramasamy brings a different perspective, grounded in a background of managing AI deployments for government clients during his six years at Amazon Web Services. His declaration disputes the government’s concerns that Anthropic could interfere with military operations. He asserts that once the Claude model is embedded in a government-secured system operated by a third-party contractor, Anthropic has no access to it, lacking any remote kill switch or mechanism to update the system without explicit Pentagon approval. He adds that Anthropic is completely blind to the operations of government users.
Addressing Concerns Over Foreign National Hiring
Ramasamy also counters the government’s assertion that employing foreign nationals equates to a security risk. He highlights that Anthropic personnel have undergone U.S. government security clearance processes, similar to those required for access to classified information. Notably, he claims that Anthropic may be the only AI firm where cleared personnel actively develop AI models designed for classified environments.
Lawsuit Allegations Against the Government
In its lawsuit, Anthropic argues that the supply-chain risk designation—the first ever applied to an American company—serves as government retaliation for the company’s outspoken stance on AI safety, infringing upon First Amendment rights. The government has filed a comprehensive response, categorically rejecting this interpretation and asserting that Anthropic’s business decision to restrict military use of its technology should not be construed as protected speech, framing the designation as a necessary national security measure rather than a punitive action.